6370
Biochemistry 2003, 42, 6370-6379
Structure of a Substrate Complex of Mammalian Cytochrome P450 2C5 at 2.3 Å Resolution: Evidence for Multiple Substrate Binding Modes†,‡ Michael R. Wester,§ Eric F. Johnson,*,§ Cristina Marques-Soares,| Patrick M. Dansette,| Daniel Mansuy,| and C. David Stout*,⊥ Department of Molecular and Experimental Medicine, The Scripps Research Institute, 10550 North Torrey Pines Road, MEM-255, La Jolla, California 92037, Laboratoire de Chimie et Biochimie Pharmacologiques et Toxicologiques, UMR 8601 CNRS, UniVersite´ Paris V45, Rue des Saints-Pe` res, 75270 Paris Cedex 06, France, and Department of Molecular Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, MB8, La Jolla, California 92037 ReceiVed December 20, 2002; ReVised Manuscript ReceiVed March 25, 2003
ABSTRACT: The structure of rabbit microsomal cytochrome P450 2C5/3LVdH complexed with a substrate, 4-methyl-N-methyl-N-(2-phenyl-2H-pyrazol-3-yl)benzenesulfonamide (DMZ), was determined by X-ray crystallography to 2.3 Å resolution. Substrate docking studies and electron density maps indicate that DMZ binds to the enzyme in two antiparallel orientations of the long axis of the substrate. One orientation places the principal site of hydroxylation, the 4-methyl group, 4.4 Å from the heme Fe, whereas the alternate conformation positions the second, infrequent site of hydroxylation at >5.9 Å from the heme Fe. Comparison of this structure to that obtained previously for the enzyme indicates that the protein closes around the substrate and prevents open access of water from bulk solvent to the heme Fe. This reflects a ∼1.5 Å movement of the F and G helices relative to helix I. The present structure provides a complete model for the protein from residues 27-488 and defines two new helices F′ and G′. The G′ helix is likely to contribute to interactions of the enzyme with membranes. The relatively large active site, as compared to the volume occupied by the substrate, and the flexibility of the enzyme are likely to underlie the capacity of drug-metabolizing enzymes to metabolize structurally diverse substrates of different sizes.
The P450 monooxygenases that metabolize xenobiotics constitute the majority of the more than 50 P450 genes in mammalian genomes. In contrast to the P450s that evolved to catalyze specialized biosynthetic pathways, the drug metabolizing P450s exhibit broad and overlapping substrate specificities. This metabolic diversity provides an effective means of oxidizing and eliminating foreign compounds such as drugs and environmental toxins. The xenobiotic metabolizing enzymes generally exhibit distinct substrate specificity profiles and often oxidize a structurally diverse range of substrates in a regiospecific manner. In some cases, multiple products arising from the oxidation of noncontiguous sites on the same substrate are obtained, suggesting multiple modes of substrate binding to the enzyme’s active site or that the mobility of substrates in the active site of the enzyme is not highly restrained. The structural basis for the metabolic diversity and selectivity exhibited by mammalian xenobiotic metabolizing P450s is poorly understood, and on the basis of current knowledge, it is difficult to predict the substrate † This investigation was supported by NIH Grants GM31001 (E.F.J.) and GM59229 (C.D.S.) and by CNRS and the French Minister of Research (D.M.). ‡ Structural coordinates have been deposited with the Protein Data Bank under accession code 1N6B. * To whom correspondence should be addressed. (E.F.J.) Tel.: (858) 784-7918. Fax: (858) 784-7978. E-mail:
[email protected]. (C.D.S.) Tel.: (858) 784-8738. Fax: (858) 784-2857. E-mail:
[email protected]. § Department of Molecular and Experimental Medicine, The Scripps Research Institute. | Universite´ Paris V45. ⊥ Department of Molecular Biology, The Scripps Research Institute.
specificities and metabolic products for these enzymes. The present study addressed whether conformational changes occur when CYP2C51 binds substrates. Crystallization of a modified form of rabbit microsomal CYP2C5 enabled determination of the first structure of a mammalian, microsomal P450 by X-ray crystallography (1) and provides a basis for experimental characterization of enzyme substrate complexes. The modified enzyme, P450 2C5/3LVdH, was produced by substitution of a short, positively charged N-terminal sequence for the native transmembrane leader sequence and by addition of a 4-histidine tag to the C-terminus. Additional amino acid substitutions that alter the surface of the protein in the vicinity of the F helix further improved the solubility and monodispersity of the enzyme in high salt buffers (2). The modified enzyme was used in this study to determine the first structure of a drug metabolizing P450 with a substrate bound in the active site. The present manuscript reports on the structure of P450 2C5/3LVdH complexed with 4-methyl-N-methyl-N-(2-phenyl-2H-pyrazol-3-yl)benzenesulfonamide (DMZ) and discusses the implications of the observed structural changes 1 Abbreviations: CVFF, consistent valence force field; CYP, cytochrome P450; DMSO, dimethyl sulfoxide; DMZ, 4-methyl-Nmethyl-N-(2-phenyl-2H-pyrazol-3-yl)benzenesulfonamide; DTT, dithiothreitol; EDTA, ethylenediamine tetraacetic acid; Fo, observed structure factor; Fc, calculated structure factor; HEPES, N-(2-hydroxyethyl)piperazine-N′-(2-ethanesulfonic acid); PDB, Protein Data Base (http:// www.rcsb.org/pdb/); rms, root-mean-square; SSRL, Stanford Synchrotron Radiation Laboratory.
10.1021/bi0273922 CCC: $25.00 © 2003 American Chemical Society Published on Web 05/08/2003
Multiple Modes of Substrate Binding to P450 2C5 with regard to the substrate selectivity of mammalian microsomal P450s. DMZ is one of several analogues of sulfaphenazole, a potent and relatively selective inhibitor of human P450 2C9, that were synthesized to examine structure/ activity relationships between the human 2C enzymes. In contrast to sulfaphenazole, DMZ proved to be a relatively general inhibitor of the human 2C enzymes (3-5). Subsequent studies described in the accompanying paper (6) showed that DMZ is a good substrate for 2C5dH, 2C5/ 3LVdH and the four human CYP 2C enzymes, and that DMZ primarily undergoes hydroxylation of its benzylic methyl substituent. On the basis of these observations, the structure of 2C5/3LVdH complexed with DMZ was determined to better understand structural determinants of substrate and inhibitor binding. EXPERIMENTAL PROCEDURES Structure Determination. DMZ was synthesized as described previously (4). P450 2C5/3LVdH was purified using a procedure that employed the detergent, CYMAL-5 (Anatrace) as described (7). Briefly, Escherichia coli expressed 2C5/3LVdH and was harvested from 3.0 L of culture by centrifugation, lysozyme treatment, and sonication. The P450 was purified by metal ion affinity column chromatography followed by CM-Sepharose ion exchange chromatography. CYMAL-5 detergent was introduced during the final wash and elution of the metal ion affinity column and was maintained in the buffers through the loading of the CM column. At this point, the column was washed extensively with detergent free buffer, and the P450 was eluted in 50 mM potassium phosphate buffer, pH 7.4 containing 500 mM NaCl, 1 mM EDTA, 0.2 mM DTT, and 20% glycerol. The purified protein was concentrated using a centrifugal concentrating device and combined with an equimolar amount of DMZ. The complex was crystallized by the vapor diffusion method using 2.5 µL hanging drops containing 0.24 mM P450, 0.24 mM DMZ, 2.4 mM CYMAL-5 detergent, 1.1 M ammonium sulfate, 0.05 M HEPES pH 7.5, 0.5% PEG400, 20 mM potassium phosphate pH 7.4, 200 mM NaCl, 0.4 mM EDTA, 0.08 mM DTT, and 8% glycerol. The drops were equilibrated against 2.2 M ammonium sulfate, containing 0.1 M HEPES, pH 7.5 and 1% PEG400 at 24 °C. Crystals were prepared for data collection by briefly soaking them in a cryoprotectant of 2.2 M ammonium sulfate, containing 0.1 M HEPES, pH 7.5, 1% PEG400, and 20% sucrose followed by flash-freezing in liquid N2 (7). Crystals were then transferred to the cryo-stream, and data were collected at Stanford Synchrotron Radiation Laboratory (SSRL) beam line 9-2. Data analyzed here were collected at 100 K using a single crystal of dimensions 0.3 × 0.5 × 0.7 mm and were recorded using a Quantum4 CCD detector and 1° oscillations (90 frames, 45 s exposure). Low-resolution, saturated reflections were collected again using 4° oscillations (25 frames, 10 s exposure) to obtain accurate intensity data and merged with the high resolution data. The crystal did not decay noticeably during data collection and exhibited uniform mosaicity of ∼0.6°. The data were processed with CCP4 programs Mosflm and Scala (8, 9), and a statistical analysis of the X-ray diffraction data is presented in Table 1. The structure of P450 2C5 3LV/dH determined at 3.0 Å (PDB code 1DT6) was used as a starting point for crystallographic refinement. The model was positioned in the unit
Biochemistry, Vol. 42, No. 21, 2003 6371 Table 1: Data Collection and Refinement Statistics P450 construct no. of crystals complex space group unit cell (a, b, c) (Å)
2C5/3LVdH 1 DMZ I222 74.33, 134.29, 171.84
Data Collection SSRL beam line BL 9-2 wavelength (Å) 0.979 resolution range (Å) 50.0-2.30 total observations 153 575 unique reflections > 0.0 σF 37 253 redundancya 4.1 (2.4) completeness %a 96.6 (78.5) 〈I/σI〉a 13.2 (1.4) Rsymm (I)a 0.058 (0.479) Refinement R-factor Rfree (5% of data) rms deviation bonds (Å) rms deviation angles (deg)b
0.257 0.292 0.008 1.29
Model residues/no. of atoms/av. B-factor (Å2) proteinc 3698 heme 43 DMZd 46 water molecules 118 sulfate ions 10
60.4 39.5 80.9 61.0 81.0
a Values for the highest resolution shell in parentheses. Values of 〈I/σI〉 < 2.0 are accepted because of some anisotropy in the diffraction. b Ramachandran plot: 85.2% of residues in most favored regions; 13.3% in allowed regions; 0.7% in generously allowed regions; and 0.7% in disfavored regions. c Residues 27-488 of the 2C5 3LV/dH construct. d Includes AC1 and AC2 conformations.
cell of the DMZ complex by rigid body refinement at 4.0, 3.5, and subsequently 3.0 Å resolutions. The model was subsequently refined against the 2.3 Å data by conjugate gradient least squares minimization, simulated annealing, and individual atomic, isotropic B-factor refinement using CNS (10). Standard refinement protocols were employed, and the refinement proceeded normally through multiple cycles alternating with interpretation, editing, and adjustment of the model into σA-weighted 2|Fo| - |Fc| composite omit and |Fo| - |Fc| electron density maps using the program Xfit/ Xtalview (11). Altogether, the model was evaluated, edited, adjusted, and rebuilt nine times into successive composite omit maps. The increased resolution of the electron density allowed several improvements in the protein model to be made while also indicating a number of conformational changes resulting from substrate binding. While many smaller changes were made, significant adjustment (but not rethreading) of the model was necessary for residues 2848 (N-terminal β-strand and turn, and A′-helix), 95-99 (Arg97 and Arg430 contacts to the heme propionates), 100108 (B′-helix), 114-117 (active site side chains), 271-277 (polar surface loop), 412-417 (β-turn and meander region), and 470-474 (β-turn in the active site). In addition, strong electron density was apparent for the region between helices F and G allowing residues 208-211 and 223-228 to be adjusted and the intervening segment of residues 212-222 to be completed. Excluding these residues, the rms deviation of 405 CR atoms between the starting model (1DT6), the substrate-free enzyme, and the refined DMZ complex (Table 1) is 0.66 Å. In the final stages, H2O molecules were included
6372 Biochemistry, Vol. 42, No. 21, 2003 in the model, refined, and edited (Table 1). Strong difference electron density was initially apparent in the active site for DMZ. This density was modeled in the final stages of refinement and interpreted in terms of two alternate conformations suggested by automated substrate-docking studies (below). The occupancy of each conformation (AC1, AC2) was set to 0.5 based on comparison of refined B-factors for the two models of the substrate versus surrounding portions of the protein. The model has good stereochemistry with only six residues in generously allowed or disallowed regions of the Ramachandran plot (Table 1). Of these, three are in weak density (Lys28, Glu272, Arg374) at the N-terminus or in loops, and three are well-defined in strong density (Ile38, Ala117, Ser426). The electron density defines the χ1 and χ2 torsion angles of most Val, Leu, and Ile side chains, whereas Thr, Asn, Gln, and His rotamers are additionally defined by hydrogen-bonding interactions. Coordinates for the DMZ complex of P450 2C5/3LVdH have been deposited in the PDB with accession code 1N6B. Automated Docking of DMZ. Computer simulated automated docking studies were performed with an explicit hydrogen model generated from the model for the protein using REDUCE (12). Preliminary docking employed AUTODOCK 3.05, a grid based docking program (13), to examine probable binding sites for DMZ using a modified genetic algorithm that employs a local search to identify low energy binding sites and orientations of the probe molecule. A 50 × 50 × 50 grid with a spacing of 0.375 Å centered at -10.11, -14.19, and -26.68 Å that encompassed the active site was used. The parameters for the HC atom type in the AMBER parameter set, parm91 (14), were used to define the Leonard-Jones 12-6 pair potentials for the nonpolar hydrogens. The initial conformations of the molecule were based on the crystal structure of sulfaphenazole (15), which exhibits both a ring stacked and an extended conformation. The structure of DMZ was built in INSIGHT II (Accelrys) by substituting methyl groups for the amide hydrogen and the aromatic amino group of sulfaphenazole. Partial charges were assigned in INSIGHT II based on the Consistent Valence Force Field (CVFF). Flexible bonds were defined in the substrate to allow internal rotations around unconjugated bonds. The results of 50 randomly seeded runs were analyzed for each of the two conformations of DMZ. The results of each run were clustered when atomic coordinates of the models exhibited less than a 1 Å root-mean-square difference from each other. The results obtained were independent of the initial conformation of the substrate model. Representative structures of the two lowest energy clusters that closely matched the electron density map were used for subsequent crystallographic refinement as alternate conformers designated AC1 and AC2. RESULTS Improvements to the purification procedures for the enzyme that included the use of the detergent CYMAL-5 led to the identification of additional crystallization conditions that improved the limits of diffraction (7). Crystals were grown in the presence of equimolar concentrations of DMZ and protein. Although the enzyme crystallized in the absence of substrate, the crystals did not diffract as well. The initial concentrations of the enzyme and substrate were 240 µM in
Wester et al. the drop, concentrations that greatly exceed the apparent binding constant of 20 µM estimated from substrate induced changes in the spin state of the enzyme. A data set collected for a crystal that diffracted to 2.3 Å resolution was used for model building and refinement. An initial model was obtained by molecular replacement and finalized by sequential rounds of building, fitting, and refinement. The final model exhibits an R value of 0.26 and an Rfree value of 0.29 (Table 1). As shown in Figure 1, the polypeptide fold for the structure of the enzyme substrate complex corresponds closely to that of the original structure, PDB code 1DT6, but significant conformation changes were observed in regions forming the distal portions of the substrate binding site that required extensive rebuilding (Experimental Procedures). These regions show larger B values than are seen for the structural core (Figure 1a). The CR atoms of the F and G helices have moved roughly 1-1.5 Å relative to the axis of helix I (Figure 1c). Helix I is relatively straight in the earlier structure of 2C5/3LVdH, but it exhibits a distinct bend that displaces the N-terminal end of the helix in the current structure (Figure 1c). The higher resolution and improved B values for the region between helices F and G, residues 205 to 230, allowed this portion of the enzyme to be modeled for the first time, Figure 2. As shown in Figures 1 and 2, the F-G loop contains two short helical segments, F′ and G′, separated by a turn. The G′ helix contributes to a hydrophobic surface on the catalytic domain near the N-terminus where it joins the transmembrane leader sequence. This region is thought to contribute to the membrane interactions exhibited by the truncated enzyme, 2C5dH (1). The G′ helix is largely hydrophobic with the exception of D224 and is flanked by aromatic residues. The helical structure of this region increases the likelihood that the hydrophobic tip of the protein can be buried in the membrane by reducing the need to hydrate the peptide backbone (16). In addition, the region between helices B and C is better ordered revealing an additional helix, B′, which is generally seen in prokaryotic P450s (Figure 1a,b). The increased order in this region may be derived, in part, from substrate interactions with the B′, F, and G helices. As a result of the increased resolution (2.3 vs. 3.0 Å) and lower B values, the overall geometry of the model is improved relative to the previously published structure. In addition to overall better stereochemistry, β-strands are also better defined revealing a β-turn in the meander region, at residues 411-414, that is not evident in prokaryotic P450s. Examination of the |Fo| - |Fc| electron density map revealed significant electron density for DMZ in the active site of P450 2C5 (Figure 3). As the density did not uniquely define the orientation of the molecule, a computational approach was taken to model the interaction of DMZ with the refined model of 2C5/3LVdH. To model substrate interactions with the protein, an explicit hydrogen model of the protein was generated using the program REDUCE (12). REDUCE evaluates the orientations of the side-chain N and O atoms of Asn and Gln residues as well as of His χ2 rotamers by selecting the best hydrogen bonding potentials while minimizing or eliminating clashes with other residues. The results of this minimization were also used to check for consistency with respect to the X-ray refinement. The
Multiple Modes of Substrate Binding to P450 2C5
Biochemistry, Vol. 42, No. 21, 2003 6373
FIGURE 1: (a) Structure of cytochrome P450 2C5/3LVdH with DMZ bound. The principle helices and NH2- and COOH-termini are labeled. The protein is colored according to average B-factor per residue: dark blue,
